Deficits in discrimination and identification characterize a range of visual impairments. Training-based improvements in visual performance provide one possible non-invasive approach for remediation. Ideally, training for remediation will learn and transfer to a range of similar stimuli and tasks. However, improvements in perceptual task performance through perceptual learning or training, and the extent of transfer to related conditions, both depend critically upon the training protocol and the mixture of stimuli and tasks being trained. The current research uses computational models of visual perceptual learning, new and extended training and testing protocols, efficient estimation methods, and empirical tests. It aims to improve our understanding of the conditions for and the limits of transfer and specificity of perceptual learning. Past research has reported strong modulation of perceptual learning for multiple tasks/stimuli and a range of effects on transfer. Here, we propose new systematic investigations of both phenomena that cry out for an integrated theoretical account of these fundamental issues in perceptual learning. In Aim 1, we develop and test a theoretical framework to understand how practice on multiple stimuli or tasks interact in perceptual learning, in some cases eliminating learning, and in other cases supporting it. In Aim 2, we develop and test a theoretical framework to understand and predict the cases in which perceptual training transfers to other stimuli, tasks, and visual locations, and those where it does not. We extend a previous computational model of perceptual learning based on partially-supervised learning algorithms, to incorporate location-independent as well as location-specific visual representations. The goal of this research program is to develop the theories and practical implementation of perceptual learning in normal populations that could contribute to translational applications to developmental learning and to ameliorative training in populations with perceptual deficits. These aims are consistent with the goals of the NEI's National Plan for Eye and Vision Research. PUBLIC HEALTH RELEVANCE: Perceptual learning through training visual tasks is one approach to remediation of some visual impairment and can contribute to development of visual skills. The current project seeks to understand the conditions for producing the best learning of multiple stimuli and tasks and for improving transfer of training to related stimuli and tasks. Through modeling learning for different training protocols, the proposed research program of model development and empirical testing aims to produce a framework for predicting the value of different training regimens in normal adults, and suggest applications in developmental and rehabilitative training.